Analysis of the algorithm: From rankings to backup genes.

Ranking section

Table 1. Performance summary for every kernel and net before integration.

Net Kernel tpr_down_ci_0.95 tpr_up_ci_0.95 fpr_down_ci_0.95 fpr_up_ci_0.95 f
gene_TF ct 1 1 0 0 1
gene_TF el 1 1 0 0 1
gene_TF ka 1 1 1 1 0.5
gene_TF node2vec 1 1 0 0 1
gene_TF rf 1 1 0 0 1
gene_hgncGroup ct 1 1 0.75 1 0.9
gene_hgncGroup el 1 1 1 1 0.9
gene_hgncGroup ka 1 1 1 1 0.9
gene_hgncGroup node2vec 1 1 1 1 0.9
gene_hgncGroup rf 1 1 1 1 0.9

Table 2. Performance summary for every kernel after integration

Integration Kernel tpr_down_ci_0.95 tpr_up_ci_0.95 fpr_down_ci_0.95 fpr_up_ci_0.95 f
integration_mean_by_presence ct 0.75 1 0 1 0.947368421052632
integration_mean_by_presence el 1 1 0 0 1
integration_mean_by_presence ka 1 1 0 0 1
integration_mean_by_presence node2vec 0.75 1 0 0.666666666666667 0.947368421052632
integration_mean_by_presence rf 0.85 1 0 1 0.974358974358974
mean ct 0.75 1 0 1 0.947368421052632
mean el 1 1 0 0 1
mean ka 0.9 1 0 0.5 0.975609756097561
mean node2vec 0.8 1 0 0.833333333333333 0.930232558139535
mean rf 0.8 1 0 0.666666666666667 0.952380952380952

Backup Pairs Coverage By Layer Non integrated

Backup Pairs Coverage By Layer Integrated

Distribution of ranks in dataset by non integrated kernels

Distribution of ranks in dataset by Integrated kernels

ROC-AUCs for Non-integrated kernel and layer

CDF plots of non integrated kernels

ROC plots of non integrated kernels